Skillv1.0.0

ClawScan security

Frustration Translator · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

SuspiciousApr 9, 2026, 3:25 AM

Verdict: suspicious
Confidence: medium
Model: gpt-5-mini
Summary: The skill's stated purpose (detect frustration and translate into actionable instructions) matches its instructions, but the runtime rules ("execute what they meant", "never add caveats", and logging for 'improvement over time') give the agent broad discretion to act and persist user content without declaring where or how — this mismatch warrants caution.
Guidance: This skill is coherent with its goal (turning frustrated messages into clearer instructions), but it gives the agent broad permission to act quickly and to persist conversation-derived logs without declaring where data will be stored. Before installing, ask: (1) Will the agent perform potentially destructive or irreversible actions without asking for confirmation? (2) Where are appended frustration logs written and who can read them? (3) Can you disable automatic execution and require explicit confirmation for sensitive actions? If you need this skill, prefer enabling it only for user-invoked sessions (not autonomous runs), require confirmations for any action with side effects, and insist on a clear data-retention policy or local-only logging to avoid unintended data persistence.

Review Dimensions

Purpose & Capability: noteName and description align with the instructions: the skill detects emotional charge and extracts intents. However, the SKILL.md and the included frustration-log imply persistent logging and interaction with other agent subsystems (references to pipelines, spawners, and autonomous execution in the log). The skill does not declare any storage/config path or credentials for such persistence, creating a mild mismatch between intended capability (learning over time) and declared requirements (none).
Instruction Scope: concernThe instructions explicitly tell the agent to "translate" and then "execute what they meant." Combined with rules like "Never slow down" and "Never add caveats when they want action," this grants wide discretion to perform tasks quickly and with fewer safety checks. The SKILL.md also instructs checking prior messages and appending to a frustration log, which implies reading conversation history and persisting derived data. Those behaviors can lead to the agent taking actions beyond explicit user confirmation or storing sensitive conversational content without a declared storage policy.
Install Mechanism: okInstruction-only skill with no install spec and no binaries or external packages requested. No files will be downloaded or installed by an installer; lowest-risk delivery mechanism from a code-execution/install perspective.
Credentials: okThe skill requests no environment variables, credentials, or config paths, which is proportionate for a text-processing translator. Note: the included log references systems (SiteBlitz, spawners, pipelines) but the skill does not request access to them — if used in contexts that require those integrations, additional permissions would be needed and should be scrutinized.
Persistence & Privilege: concernThe SKILL.md and references/frustration-log.md encourage appending detected events for "dream review and pattern detection," implying persistent storage of user messages and derived intent mappings. The skill bundle contains a log file template but does not declare where runtime writes would go or whether persisted data is shared externally. This lack of declared persistence or retention policy is a concern for user privacy and data governance.